This notebook explores stock market data, focusing on technology giants like Apple, Amazon, Google, and Microsoft. It demonstrates the use of yfinance to retrieve stock information and visualization techniques with Seaborn and Matplotlib. The analysis includes assessing stock risk using historical performance data and predicting future prices using a Linear Regression model.
# Install necessary libraries
!pip install -q yfinance
!pip install pandas-datareader
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from pandas_datareader.data import DataReader
import yfinance as yf
from pandas_datareader import data as pdr
from datetime import datetime
yf.pdr_override()
# Set plotting styles
sns.set_style('whitegrid')
plt.style.use("fivethirtyeight")
%matplotlib inline
[notice] A new release of pip is available: 23.2.1 -> 24.0 [notice] To update, run: python.exe -m pip install --upgrade pip
Requirement already satisfied: pandas-datareader in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (0.10.0) Requirement already satisfied: lxml in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from pandas-datareader) (5.2.2) Requirement already satisfied: pandas>=0.23 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from pandas-datareader) (2.0.1) Requirement already satisfied: requests>=2.19.0 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from pandas-datareader) (2.31.0) Requirement already satisfied: python-dateutil>=2.8.2 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from pandas>=0.23->pandas-datareader) (2.9.0.post0) Requirement already satisfied: pytz>=2020.1 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from pandas>=0.23->pandas-datareader) (2024.1) Requirement already satisfied: tzdata>=2022.1 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from pandas>=0.23->pandas-datareader) (2024.1) Requirement already satisfied: numpy>=1.21.0 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from pandas>=0.23->pandas-datareader) (1.23.5) Requirement already satisfied: charset-normalizer<4,>=2 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from requests>=2.19.0->pandas-datareader) (3.3.2) Requirement already satisfied: idna<4,>=2.5 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from requests>=2.19.0->pandas-datareader) (3.6) Requirement already satisfied: urllib3<3,>=1.21.1 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from requests>=2.19.0->pandas-datareader) (2.2.1) Requirement already satisfied: certifi>=2017.4.17 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from requests>=2.19.0->pandas-datareader) (2024.2.2) Requirement already satisfied: six>=1.5 in c:\users\admin\documents\mlai\venv\venv\tensorflow_cpu\lib\site-packages (from python-dateutil>=2.8.2->pandas>=0.23->pandas-datareader) (1.16.0)
[notice] A new release of pip is available: 23.2.1 -> 24.0 [notice] To update, run: python.exe -m pip install --upgrade pip
import pandas as pd
from datetime import datetime
#tech_list = ['AAPL', 'GOOG', 'MSFT', 'AMZN']
tech_list = ['JPMorgan', 'Bank of America', 'Wells FargoC', 'Citigroup']
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# Download stock data
JPM = yf.download('JPM', start=start, end=end)
BAC = yf.download('BAC', start=start, end=end)
WFC = yf.download('WFC', start=start, end=end)
C = yf.download('C', start=start, end=end)
company_list = [JPM, BAC, WFC, C]
company_name = ['JPMorgan', 'Bank of America', 'Wells FargoC', 'Citigroup']
# for stock in tech_list:
# globals()[stock] = yf.download(stock, start=start, end=end)
[*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed
# Fill missing values using forward fill
for company in company_list:
company.ffill(inplace=True) # Forward fill to handle missing data
# Ensure consistent date format
for company in company_list:
company.reset_index(inplace=True) # Reset index to ensure consistent date format
company['Date'] = pd.to_datetime(company['Date']) # Ensure 'Date' column is in datetime format
company.set_index('Date', inplace=True) # Set 'Date' as index again
# Add company name column to each dataframe
for company, com_name in zip(company_list, company_name):
company["company_name"] = com_name
# Concatenate individual stock data into a single DataFrame
df = pd.concat(company_list, axis=0)
# Shuffle the data and get a random sample of the last 10 rows
df = df.sample(frac=1).reset_index(drop=True) # Shuffle the data
print(df.tail(10))
# Select the last 10 rows after shuffling
last_10_rows = df.tail(10)
# Print the last 10 rows with company names on the same line
print("Displaying the last 10 rows of the shuffled DataFrame:\n")
for index, row in last_10_rows.iterrows():
print(f"Company: {row['company_name']}, Date: {row.name}, Open: {row['Open']:.2f}, High: {row['High']:.2f}, "
f"Low: {row['Low']:.2f}, Close: {row['Close']:.2f}, Adj Close: {row['Adj Close']:.2f}, "
f"Volume: {row['Volume']}")# Display the last 10 rows of the shuffled dataframe
Open High Low Close Adj Close Volume
998 171.619995 171.789993 170.100006 170.660004 169.670303 9923600 \
999 36.930000 37.599998 36.880001 37.509998 37.283237 43219500
1000 53.540001 53.650002 52.990002 53.259998 52.302116 14402200
1001 46.770000 47.060001 46.220001 46.820000 46.493435 25209900
1002 142.190002 143.770004 141.850006 143.000000 141.299393 9533100
1003 28.760000 29.129999 28.760000 28.980000 28.376829 35192000
1004 174.779999 176.809998 173.699997 175.789993 174.770538 8539300
1005 34.889999 35.750000 34.750000 35.230000 35.017021 65951400
1006 163.080002 165.279999 162.389999 165.229996 163.265030 20305900
1007 155.820007 157.860001 155.820007 156.839996 154.974808 8913000
company_name
998 JPMorgan
999 Bank of America
1000 Citigroup
1001 Wells FargoC
1002 JPMorgan
1003 Bank of America
1004 JPMorgan
1005 Bank of America
1006 JPMorgan
1007 JPMorgan
Displaying the last 10 rows of the shuffled DataFrame:
Company: JPMorgan, Date: 998, Open: 171.62, High: 171.79, Low: 170.10, Close: 170.66, Adj Close: 169.67, Volume: 9923600
Company: Bank of America, Date: 999, Open: 36.93, High: 37.60, Low: 36.88, Close: 37.51, Adj Close: 37.28, Volume: 43219500
Company: Citigroup, Date: 1000, Open: 53.54, High: 53.65, Low: 52.99, Close: 53.26, Adj Close: 52.30, Volume: 14402200
Company: Wells FargoC, Date: 1001, Open: 46.77, High: 47.06, Low: 46.22, Close: 46.82, Adj Close: 46.49, Volume: 25209900
Company: JPMorgan, Date: 1002, Open: 142.19, High: 143.77, Low: 141.85, Close: 143.00, Adj Close: 141.30, Volume: 9533100
Company: Bank of America, Date: 1003, Open: 28.76, High: 29.13, Low: 28.76, Close: 28.98, Adj Close: 28.38, Volume: 35192000
Company: JPMorgan, Date: 1004, Open: 174.78, High: 176.81, Low: 173.70, Close: 175.79, Adj Close: 174.77, Volume: 8539300
Company: Bank of America, Date: 1005, Open: 34.89, High: 35.75, Low: 34.75, Close: 35.23, Adj Close: 35.02, Volume: 65951400
Company: JPMorgan, Date: 1006, Open: 163.08, High: 165.28, Low: 162.39, Close: 165.23, Adj Close: 163.27, Volume: 20305900
Company: JPMorgan, Date: 1007, Open: 155.82, High: 157.86, Low: 155.82, Close: 156.84, Adj Close: 154.97, Volume: 8913000
df = df.reset_index()
df = df.fillna(method='ffill')
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style('whitegrid')
plt.style.use("fivethirtyeight")
plt.figure(figsize=(15, 10))
plt.subplots_adjust(top=1.25, bottom=1.2)
for i, company in enumerate(company_list, 1):
plt.subplot(2, 2, i)
company['Adj Close'].plot()
plt.ylabel('Adj Close')
plt.title(f"Closing Price of {tech_list[i - 1]}")
plt.tight_layout()
plt.figure(figsize=(15, 10))
plt.subplots_adjust(top=1.25, bottom=1.2)
for i, company in enumerate(company_list, 1):
plt.subplot(2, 2, i)
company['Volume'].plot()
plt.ylabel('Volume')
plt.title(f"Sales Volume for {tech_list[i - 1]}")
plt.tight_layout()
ma_day = [10, 20, 50]
for ma in ma_day:
for company in company_list:
column_name = f"MA for {ma} days"
company[column_name] = company['Adj Close'].rolling(ma).mean()
fig, axes = plt.subplots(nrows=2, ncols=2, figsize=(15, 10))
JPM[['Adj Close', 'MA for 10 days', 'MA for 20 days', 'MA for 50 days']].plot(ax=axes[0,0])
axes[0,0].set_title('JPMORGAN')
BAC[['Adj Close', 'MA for 10 days', 'MA for 20 days', 'MA for 50 days']].plot(ax=axes[0,1])
axes[0,1].set_title('BANK OF AMERICA')
WFC[['Adj Close', 'MA for 10 days', 'MA for 20 days', 'MA for 50 days']].plot(ax=axes[1,0])
axes[1,0].set_title('Wells Fargo')
C[['Adj Close', 'MA for 10 days', 'MA for 20 days', 'MA for 50 days']].plot(ax=axes[1,1])
axes[1,1].set_title('Citigroup')
fig.tight_layout()
for company in company_list:
company['Daily Return'] = company['Adj Close'].pct_change()
fig, axes = plt.subplots(nrows=2, ncols=2, figsize=(15, 10))
JPM['Daily Return'].plot(ax=axes[0,0], legend=True, linestyle='--', marker='o')
axes[0,0].set_title('JPMORGAN')
BAC['Daily Return'].plot(ax=axes[0,1], legend=True, linestyle='--', marker='o')
axes[0,1].set_title('BANK OF AMERICA')
WFC['Daily Return'].plot(ax=axes[1,0], legend=True, linestyle='--', marker='o')
axes[1,0].set_title('Wells Fargo & Company')
C['Daily Return'].plot(ax=axes[1,1], legend=True, linestyle='--', marker='o')
axes[1,1].set_title('Citigroup Inc.')
fig.tight_layout()
plt.figure(figsize=(12, 9))
for i, company in enumerate(company_list, 1):
plt.subplot(2, 2, i)
company['Daily Return'].hist(bins=50)
plt.xlabel('Daily Return')
plt.title(f'{company_name[i - 1]}')
plt.tight_layout()
import yfinance as yf
import pandas as pd
import matplotlib.pyplot as plt
# Define company tickers
tech_list = ['JPM', 'BAC', 'WFC', 'C']
# Download stock data for the past year
end = pd.to_datetime('2024-05-27')
start = end - pd.DateOffset(years=1)
company_list = []
for stock in tech_list:
company_list.append(yf.download(stock, start=start, end=end))
# Ensure consistent date format
for company in company_list:
company.reset_index(inplace=True) # Reset index to ensure consistent date format
company['Date'] = pd.to_datetime(company['Date']) # Ensure 'Date' column is in datetime format
company.set_index('Date', inplace=True) # Set 'Date' as index again
# Calculate daily percentage change
for company in company_list:
company['Daily Change'] = company['Adj Close'].pct_change() * 100
# Create histogram (one plot for all companies)
plt.figure(figsize=(12, 6))
for i, company in enumerate(company_list):
# Overlay histograms on the same plot with different transparency levels
plt.hist(company['Daily Change'], bins=20, alpha=0.7, label=tech_list[i])
plt.xlabel('Daily Change (%)')
plt.ylabel('Frequency')
plt.title('Distribution of Daily Change for Tech Stocks (Past Year)')
plt.legend() # Add legend to identify each company's distribution
plt.tight_layout()
# ... (Rest of the code for identifying best/worst performers and displaying additional info)
plt.show()
[*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed
closing_df = pdr.get_data_yahoo(tech_list, start=start, end=end)['Adj Close']
tech_rets = closing_df.pct_change()
sns.pairplot(tech_rets, kind='reg')
plt.figure(figsize=(12, 10))
plt.subplot(2, 2, 1)
sns.heatmap(tech_rets.corr(), annot=True, cmap='summer')
plt.title('Correlation of stock return')
plt.subplot(2, 2, 2)
sns.heatmap(closing_df.corr(), annot=True, cmap='summer')
plt.title('Correlation of stock closing price')
[*********************100%%**********************] 4 of 4 completed
Text(0.5, 1.0, 'Correlation of stock closing price')
from sklearn.model_selection import train_test_split
# Drop missing values from both tech_rets and closing_df
tech_rets_cleaned = tech_rets.dropna()
closing_df_cleaned = closing_df.dropna()
# Align the number of rows in tech_rets_cleaned and closing_df_cleaned
min_rows = min(tech_rets_cleaned.shape[0], closing_df_cleaned.shape[0])
tech_rets_cleaned = tech_rets_cleaned[:min_rows]
closing_df_cleaned = closing_df_cleaned[:min_rows]
# Create X and y after handling missing values
X = tech_rets_cleaned.values
y = closing_df_cleaned.values
# Print the shapes of X and y for verification
print("Shape of X:", X.shape)
print("Shape of y:", y.shape)
# Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
Shape of X: (249, 4) Shape of y: (249, 4)
import numpy as np
import matplotlib.pyplot as plt
# Define colors for the histograms
train_colors = ['blue', 'green', 'orange', 'purple'] # Adjust as needed
test_colors = ['red', 'yellow', 'cyan', 'magenta'] # Adjust as needed
# Plot the distribution of y_train
plt.figure(figsize=(12, 6))
for i, dataset in enumerate(y_train.T):
stock_name = tech_list[i] # Assuming tech_list contains the names of the stocks
plt.hist(dataset, bins=30, color=train_colors[i], alpha=0.7, label=stock_name)
plt.title('Distribution of y_train')
plt.xlabel('Values')
plt.ylabel('Frequency')
plt.legend()
plt.show()
# Plot the distribution of y_test
plt.figure(figsize=(12, 6))
for i, dataset in enumerate(y_test.T):
stock_name = tech_list[i] # Assuming tech_list contains the names of the stocks
plt.hist(dataset, bins=30, color=test_colors[i], alpha=0.7, label=stock_name)
plt.title('Distribution of y_test')
plt.xlabel('Values')
plt.ylabel('Frequency')
plt.legend()
plt.show()
In this notebook, we delved into the world of stock market data analysis. Here's a summary of what we explored:
If you have any questions or need further clarification on any topic covered in this notebook, feel free to ask in the comments below. I'll be happy to assist you!
References:
import numpy as np
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler
# Assuming X_train and X_test are your input data arrays
# and y_train, y_test are corresponding target arrays
# Initialize the scaler
scaler = MinMaxScaler()
# Fit the scaler on the training data and transform it
X_train_normalized = scaler.fit_transform(X_train)
X_test_normalized = scaler.transform(X_test)
# Reshape the data for LSTM input
X_train_lstm = np.reshape(X_train_normalized, (X_train_normalized.shape[0], X_train_normalized.shape[1], 1))
X_test_lstm = np.reshape(X_test_normalized, (X_test_normalized.shape[0], X_test_normalized.shape[1], 1))
# Define the LSTM model
model_lstm = Sequential()
model_lstm.add(LSTM(units=50, return_sequences=True, input_shape=(X_train_lstm.shape[1], 1)))
model_lstm.add(LSTM(units=50, return_sequences=False))
model_lstm.add(Dense(units=1))
# Compile the model
model_lstm.compile(optimizer='adam', loss='mean_squared_error')
# Train the model with validation split
history = model_lstm.fit(X_train_lstm, y_train, epochs=100, batch_size=32, validation_split=0.2)
# Make predictions
predictions_lstm = model_lstm.predict(X_test_lstm)
# Evaluate model performance on test set if needed
# test_loss = model_lstm.evaluate(X_test_lstm, y_test)
# Print predictions or further analysis
print(predictions_lstm)
# Ensure y_train and y_test are reshaped to be one-dimensional if only one target is required
# For example, selecting the first column if it's appropriate
y_train_single = y_train[:, 0] # Or any other column/aggregation you need
y_test_single = y_test[:, 0]
# Train the model
history = model_lstm.fit(X_train_lstm, y_train_single, epochs=100, batch_size=32, validation_split=0.2)
# Make predictions
predictions_lstm = model_lstm.predict(X_test_lstm)
# Calculate metrics
mse = mean_squared_error(y_test_single, predictions_lstm)
mae = mean_absolute_error(y_test_single, predictions_lstm)
print(f'Mean Squared Error (MSE): {mse}')
print(f'Mean Absolute Error (MAE): {mae}')
Epoch 1/100 5/5 [==============================] - 4s 180ms/step - loss: 8143.8276 - val_loss: 8128.0859 Epoch 2/100 5/5 [==============================] - 0s 9ms/step - loss: 8129.9253 - val_loss: 8112.0767 Epoch 3/100 5/5 [==============================] - 0s 9ms/step - loss: 8112.1118 - val_loss: 8089.6943 Epoch 4/100 5/5 [==============================] - 0s 9ms/step - loss: 8085.9512 - val_loss: 8055.0171 Epoch 5/100 5/5 [==============================] - 0s 9ms/step - loss: 8044.2656 - val_loss: 7997.8540 Epoch 6/100 5/5 [==============================] - 0s 9ms/step - loss: 7974.5967 - val_loss: 7900.4688 Epoch 7/100 5/5 [==============================] - 0s 9ms/step - loss: 7855.5166 - val_loss: 7735.8779 Epoch 8/100 5/5 [==============================] - 0s 9ms/step - loss: 7658.1421 - val_loss: 7483.6436 Epoch 9/100 5/5 [==============================] - 0s 10ms/step - loss: 7375.0718 - val_loss: 7170.9312 Epoch 10/100 5/5 [==============================] - 0s 9ms/step - loss: 7058.8789 - val_loss: 6874.3228 Epoch 11/100 5/5 [==============================] - 0s 10ms/step - loss: 6781.7974 - val_loss: 6656.5298 Epoch 12/100 5/5 [==============================] - 0s 10ms/step - loss: 6597.1289 - val_loss: 6519.8096 Epoch 13/100 5/5 [==============================] - 0s 9ms/step - loss: 6481.7944 - val_loss: 6432.8203 Epoch 14/100 5/5 [==============================] - 0s 11ms/step - loss: 6405.0117 - val_loss: 6368.7471 Epoch 15/100 5/5 [==============================] - 0s 10ms/step - loss: 6345.2656 - val_loss: 6315.1108 Epoch 16/100 5/5 [==============================] - 0s 10ms/step - loss: 6294.5776 - val_loss: 6266.6958 Epoch 17/100 5/5 [==============================] - 0s 9ms/step - loss: 6247.1685 - val_loss: 6221.8374 Epoch 18/100 5/5 [==============================] - 0s 9ms/step - loss: 6203.2266 - val_loss: 6179.3042 Epoch 19/100 5/5 [==============================] - 0s 10ms/step - loss: 6161.3018 - val_loss: 6138.6182 Epoch 20/100 5/5 [==============================] - 0s 10ms/step - loss: 6121.4141 - val_loss: 6099.3032 Epoch 21/100 5/5 [==============================] - 0s 10ms/step - loss: 6082.2202 - val_loss: 6061.3950 Epoch 22/100 5/5 [==============================] - 0s 12ms/step - loss: 6044.6079 - val_loss: 6024.5557 Epoch 23/100 5/5 [==============================] - 0s 10ms/step - loss: 6007.8545 - val_loss: 5988.7119 Epoch 24/100 5/5 [==============================] - 0s 10ms/step - loss: 5972.1025 - val_loss: 5953.7139 Epoch 25/100 5/5 [==============================] - 0s 9ms/step - loss: 5937.5112 - val_loss: 5919.3384 Epoch 26/100 5/5 [==============================] - 0s 9ms/step - loss: 5903.0161 - val_loss: 5885.7969 Epoch 27/100 5/5 [==============================] - 0s 9ms/step - loss: 5869.6401 - val_loss: 5852.8311 Epoch 28/100 5/5 [==============================] - 0s 9ms/step - loss: 5836.6182 - val_loss: 5820.5088 Epoch 29/100 5/5 [==============================] - 0s 9ms/step - loss: 5804.5332 - val_loss: 5788.6323 Epoch 30/100 5/5 [==============================] - 0s 9ms/step - loss: 5772.7446 - val_loss: 5757.2817 Epoch 31/100 5/5 [==============================] - 0s 10ms/step - loss: 5741.6025 - val_loss: 5726.3696 Epoch 32/100 5/5 [==============================] - 0s 9ms/step - loss: 5710.4766 - val_loss: 5696.0684 Epoch 33/100 5/5 [==============================] - 0s 9ms/step - loss: 5680.2368 - val_loss: 5666.1553 Epoch 34/100 5/5 [==============================] - 0s 10ms/step - loss: 5650.2773 - val_loss: 5636.6719 Epoch 35/100 5/5 [==============================] - 0s 17ms/step - loss: 5620.7690 - val_loss: 5607.5811 Epoch 36/100 5/5 [==============================] - 0s 12ms/step - loss: 5591.7075 - val_loss: 5578.8447 Epoch 37/100 5/5 [==============================] - 0s 10ms/step - loss: 5563.0527 - val_loss: 5550.4297 Epoch 38/100 5/5 [==============================] - 0s 9ms/step - loss: 5534.6187 - val_loss: 5522.3940 Epoch 39/100 5/5 [==============================] - 0s 9ms/step - loss: 5506.5552 - val_loss: 5494.7153 Epoch 40/100 5/5 [==============================] - 0s 9ms/step - loss: 5479.0654 - val_loss: 5467.2749 Epoch 41/100 5/5 [==============================] - 0s 9ms/step - loss: 5451.4707 - val_loss: 5440.2603 Epoch 42/100 5/5 [==============================] - 0s 9ms/step - loss: 5424.3677 - val_loss: 5413.5752 Epoch 43/100 5/5 [==============================] - 0s 9ms/step - loss: 5397.8096 - val_loss: 5387.1016 Epoch 44/100 5/5 [==============================] - 0s 9ms/step - loss: 5371.2231 - val_loss: 5360.9849 Epoch 45/100 5/5 [==============================] - 0s 9ms/step - loss: 5345.0225 - val_loss: 5335.1670 Epoch 46/100 5/5 [==============================] - 0s 9ms/step - loss: 5319.1479 - val_loss: 5309.6196 Epoch 47/100 5/5 [==============================] - 0s 10ms/step - loss: 5293.5903 - val_loss: 5284.3164 Epoch 48/100 5/5 [==============================] - 0s 9ms/step - loss: 5268.1704 - val_loss: 5259.3135 Epoch 49/100 5/5 [==============================] - 0s 9ms/step - loss: 5243.1924 - val_loss: 5234.5225 Epoch 50/100 5/5 [==============================] - 0s 17ms/step - loss: 5218.2632 - val_loss: 5210.0342 Epoch 51/100 5/5 [==============================] - 0s 10ms/step - loss: 5193.8901 - val_loss: 5185.7017 Epoch 52/100 5/5 [==============================] - 0s 9ms/step - loss: 5169.4272 - val_loss: 5161.6797 Epoch 53/100 5/5 [==============================] - 0s 9ms/step - loss: 5145.4585 - val_loss: 5137.8477 Epoch 54/100 5/5 [==============================] - 0s 9ms/step - loss: 5121.6353 - val_loss: 5114.2490 Epoch 55/100 5/5 [==============================] - 0s 9ms/step - loss: 5097.9087 - val_loss: 5090.9404 Epoch 56/100 5/5 [==============================] - 0s 10ms/step - loss: 5074.3457 - val_loss: 5067.9478 Epoch 57/100 5/5 [==============================] - 0s 9ms/step - loss: 5051.5513 - val_loss: 5045.0117 Epoch 58/100 5/5 [==============================] - 0s 9ms/step - loss: 5028.5298 - val_loss: 5022.3442 Epoch 59/100 5/5 [==============================] - 0s 9ms/step - loss: 5005.6895 - val_loss: 4999.9575 Epoch 60/100 5/5 [==============================] - 0s 9ms/step - loss: 4983.5366 - val_loss: 4977.6328 Epoch 61/100 5/5 [==============================] - 0s 9ms/step - loss: 4960.8755 - val_loss: 4955.7051 Epoch 62/100 5/5 [==============================] - 0s 9ms/step - loss: 4938.9102 - val_loss: 4933.9351 Epoch 63/100 5/5 [==============================] - 0s 9ms/step - loss: 4917.0591 - val_loss: 4912.3662 Epoch 64/100 5/5 [==============================] - 0s 9ms/step - loss: 4895.4795 - val_loss: 4890.9648 Epoch 65/100 5/5 [==============================] - 0s 9ms/step - loss: 4874.0874 - val_loss: 4869.7393 Epoch 66/100 5/5 [==============================] - 0s 9ms/step - loss: 4852.7178 - val_loss: 4848.7710 Epoch 67/100 5/5 [==============================] - 0s 9ms/step - loss: 4831.6533 - val_loss: 4828.0034 Epoch 68/100 5/5 [==============================] - 0s 9ms/step - loss: 4810.9199 - val_loss: 4807.3711 Epoch 69/100 5/5 [==============================] - 0s 14ms/step - loss: 4790.1260 - val_loss: 4786.9917 Epoch 70/100 5/5 [==============================] - 0s 10ms/step - loss: 4769.7515 - val_loss: 4766.7510 Epoch 71/100 5/5 [==============================] - 0s 9ms/step - loss: 4749.4482 - val_loss: 4746.7056 Epoch 72/100 5/5 [==============================] - 0s 9ms/step - loss: 4729.3008 - val_loss: 4726.8638 Epoch 73/100 5/5 [==============================] - 0s 9ms/step - loss: 4709.4165 - val_loss: 4707.1855 Epoch 74/100 5/5 [==============================] - 0s 9ms/step - loss: 4689.6343 - val_loss: 4687.7100 Epoch 75/100 5/5 [==============================] - 0s 9ms/step - loss: 4670.1152 - val_loss: 4668.3950 Epoch 76/100 5/5 [==============================] - 0s 9ms/step - loss: 4650.7793 - val_loss: 4649.2393 Epoch 77/100 5/5 [==============================] - 0s 9ms/step - loss: 4631.6421 - val_loss: 4630.2280 Epoch 78/100 5/5 [==============================] - 0s 9ms/step - loss: 4612.5049 - val_loss: 4611.4482 Epoch 79/100 5/5 [==============================] - 0s 10ms/step - loss: 4593.7314 - val_loss: 4592.8037 Epoch 80/100 5/5 [==============================] - 0s 9ms/step - loss: 4574.8765 - val_loss: 4574.4111 Epoch 81/100 5/5 [==============================] - 0s 9ms/step - loss: 4556.4604 - val_loss: 4556.1392 Epoch 82/100 5/5 [==============================] - 0s 9ms/step - loss: 4538.1118 - val_loss: 4538.0342 Epoch 83/100 5/5 [==============================] - 0s 9ms/step - loss: 4519.9336 - val_loss: 4520.0928 Epoch 84/100 5/5 [==============================] - 0s 9ms/step - loss: 4502.0205 - val_loss: 4502.2681 Epoch 85/100 5/5 [==============================] - 0s 9ms/step - loss: 4484.0474 - val_loss: 4484.6670 Epoch 86/100 5/5 [==============================] - 0s 9ms/step - loss: 4466.5254 - val_loss: 4467.1494 Epoch 87/100 5/5 [==============================] - 0s 9ms/step - loss: 4448.8374 - val_loss: 4449.8740 Epoch 88/100 5/5 [==============================] - 0s 15ms/step - loss: 4431.4033 - val_loss: 4432.7881 Epoch 89/100 5/5 [==============================] - 0s 10ms/step - loss: 4414.3193 - val_loss: 4415.7969 Epoch 90/100 5/5 [==============================] - 0s 9ms/step - loss: 4397.1855 - val_loss: 4399.0010 Epoch 91/100 5/5 [==============================] - 0s 9ms/step - loss: 4380.4302 - val_loss: 4382.2900 Epoch 92/100 5/5 [==============================] - 0s 9ms/step - loss: 4363.6353 - val_loss: 4365.7578 Epoch 93/100 5/5 [==============================] - 0s 9ms/step - loss: 4347.0581 - val_loss: 4349.3721 Epoch 94/100 5/5 [==============================] - 0s 9ms/step - loss: 4330.6431 - val_loss: 4333.1274 Epoch 95/100 5/5 [==============================] - 0s 9ms/step - loss: 4314.3701 - val_loss: 4317.0264 Epoch 96/100 5/5 [==============================] - 0s 9ms/step - loss: 4298.1382 - val_loss: 4301.1221 Epoch 97/100 5/5 [==============================] - 0s 8ms/step - loss: 4282.2021 - val_loss: 4285.3389 Epoch 98/100 5/5 [==============================] - 0s 9ms/step - loss: 4266.2593 - val_loss: 4269.7510 Epoch 99/100 5/5 [==============================] - 0s 9ms/step - loss: 4250.6953 - val_loss: 4254.2422 Epoch 100/100 5/5 [==============================] - 0s 9ms/step - loss: 4235.1309 - val_loss: 4238.8843 WARNING:tensorflow:6 out of the last 11 calls to <function Model.make_predict_function.<locals>.predict_function at 0x0000017D430BF600> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for more details. 2/2 [==============================] - 1s 3ms/step [[35.94997 ] [35.88243 ] [35.870094] [35.89025 ] [35.878727] [35.89504 ] [35.89878 ] [35.894253] [35.909214] [35.785507] [35.885708] [35.90412 ] [35.897495] [35.88516 ] [35.880836] [35.933495] [35.905518] [35.846313] [35.906784] [35.818306] [35.89645 ] [35.88505 ] [35.906113] [35.87524 ] [35.892036] [35.89049 ] [35.88505 ] [35.876785] [35.839222] [35.89672 ] [35.919434] [35.927963] [35.90435 ] [35.89111 ] [35.909206] [35.885624] [35.90399 ] [35.906624] [35.90254 ] [35.875717] [35.890366] [35.924232] [35.943573] [35.90839 ] [35.924572] [35.87931 ] [35.89358 ] [35.906345] [35.89735 ] [35.908398]] Epoch 1/100 5/5 [==============================] - 3s 299ms/step - loss: 36.0106 - val_loss: 42.2561 Epoch 2/100 5/5 [==============================] - 0s 10ms/step - loss: 37.0994 - val_loss: 43.0309 Epoch 3/100 5/5 [==============================] - 0s 9ms/step - loss: 37.6379 - val_loss: 43.3818 Epoch 4/100 5/5 [==============================] - 0s 10ms/step - loss: 37.8643 - val_loss: 43.4786 Epoch 5/100 5/5 [==============================] - 0s 10ms/step - loss: 37.8914 - val_loss: 43.4187 Epoch 6/100 5/5 [==============================] - 0s 11ms/step - loss: 37.8016 - val_loss: 43.2673 Epoch 7/100 5/5 [==============================] - 0s 10ms/step - loss: 37.6352 - val_loss: 43.0674 Epoch 8/100 5/5 [==============================] - 0s 9ms/step - loss: 37.4332 - val_loss: 42.8354 Epoch 9/100 5/5 [==============================] - 0s 9ms/step - loss: 37.2116 - val_loss: 42.5823 Epoch 10/100 5/5 [==============================] - 0s 9ms/step - loss: 36.9725 - val_loss: 42.3185 Epoch 11/100 5/5 [==============================] - 0s 8ms/step - loss: 36.7285 - val_loss: 42.0474 Epoch 12/100 5/5 [==============================] - 0s 9ms/step - loss: 36.4697 - val_loss: 41.7779 Epoch 13/100 5/5 [==============================] - 0s 9ms/step - loss: 36.2144 - val_loss: 41.5089 Epoch 14/100 5/5 [==============================] - 0s 9ms/step - loss: 35.9682 - val_loss: 41.2362 Epoch 15/100 5/5 [==============================] - 0s 15ms/step - loss: 35.7135 - val_loss: 40.9655 Epoch 16/100 5/5 [==============================] - 0s 11ms/step - loss: 35.4615 - val_loss: 40.6962 Epoch 17/100 5/5 [==============================] - 0s 9ms/step - loss: 35.2054 - val_loss: 40.4316 Epoch 18/100 5/5 [==============================] - 0s 9ms/step - loss: 34.9553 - val_loss: 40.1691 Epoch 19/100 5/5 [==============================] - 0s 9ms/step - loss: 34.7189 - val_loss: 39.9024 Epoch 20/100 5/5 [==============================] - 0s 9ms/step - loss: 34.4687 - val_loss: 39.6400 Epoch 21/100 5/5 [==============================] - 0s 9ms/step - loss: 34.2268 - val_loss: 39.3782 Epoch 22/100 5/5 [==============================] - 0s 9ms/step - loss: 33.9825 - val_loss: 39.1192 Epoch 23/100 5/5 [==============================] - 0s 9ms/step - loss: 33.7452 - val_loss: 38.8603 Epoch 24/100 5/5 [==============================] - 0s 9ms/step - loss: 33.5031 - val_loss: 38.6052 Epoch 25/100 5/5 [==============================] - 0s 9ms/step - loss: 33.2725 - val_loss: 38.3486 Epoch 26/100 5/5 [==============================] - 0s 9ms/step - loss: 33.0451 - val_loss: 38.0903 Epoch 27/100 5/5 [==============================] - 0s 9ms/step - loss: 32.7983 - val_loss: 37.8412 Epoch 28/100 5/5 [==============================] - 0s 9ms/step - loss: 32.5649 - val_loss: 37.5953 Epoch 29/100 5/5 [==============================] - 0s 9ms/step - loss: 32.3381 - val_loss: 37.3500 Epoch 30/100 5/5 [==============================] - 0s 9ms/step - loss: 32.1083 - val_loss: 37.1078 Epoch 31/100 5/5 [==============================] - 0s 9ms/step - loss: 31.9006 - val_loss: 36.8583 Epoch 32/100 5/5 [==============================] - 0s 9ms/step - loss: 31.6540 - val_loss: 36.6230 Epoch 33/100 5/5 [==============================] - 0s 9ms/step - loss: 31.4403 - val_loss: 36.3851 Epoch 34/100 5/5 [==============================] - 0s 16ms/step - loss: 31.2202 - val_loss: 36.1492 Epoch 35/100 5/5 [==============================] - 0s 10ms/step - loss: 31.0061 - val_loss: 35.9131 Epoch 36/100 5/5 [==============================] - 0s 11ms/step - loss: 30.7923 - val_loss: 35.6777 Epoch 37/100 5/5 [==============================] - 0s 9ms/step - loss: 30.5695 - val_loss: 35.4491 Epoch 38/100 5/5 [==============================] - 0s 9ms/step - loss: 30.3652 - val_loss: 35.2180 Epoch 39/100 5/5 [==============================] - 0s 9ms/step - loss: 30.1481 - val_loss: 34.9923 Epoch 40/100 5/5 [==============================] - 0s 9ms/step - loss: 29.9434 - val_loss: 34.7665 Epoch 41/100 5/5 [==============================] - 0s 9ms/step - loss: 29.7385 - val_loss: 34.5417 Epoch 42/100 5/5 [==============================] - 0s 9ms/step - loss: 29.5308 - val_loss: 34.3205 Epoch 43/100 5/5 [==============================] - 0s 9ms/step - loss: 29.3301 - val_loss: 34.1003 Epoch 44/100 5/5 [==============================] - 0s 9ms/step - loss: 29.1337 - val_loss: 33.8799 Epoch 45/100 5/5 [==============================] - 0s 9ms/step - loss: 28.9263 - val_loss: 33.6661 Epoch 46/100 5/5 [==============================] - 0s 9ms/step - loss: 28.7330 - val_loss: 33.4523 Epoch 47/100 5/5 [==============================] - 0s 9ms/step - loss: 28.5350 - val_loss: 33.2421 Epoch 48/100 5/5 [==============================] - 0s 9ms/step - loss: 28.3494 - val_loss: 33.0299 Epoch 49/100 5/5 [==============================] - 0s 9ms/step - loss: 28.1567 - val_loss: 32.8208 Epoch 50/100 5/5 [==============================] - 0s 9ms/step - loss: 27.9580 - val_loss: 32.6186 Epoch 51/100 5/5 [==============================] - 0s 15ms/step - loss: 27.7809 - val_loss: 32.4128 Epoch 52/100 5/5 [==============================] - 0s 11ms/step - loss: 27.5861 - val_loss: 32.2140 Epoch 53/100 5/5 [==============================] - 0s 10ms/step - loss: 27.4054 - val_loss: 32.0156 Epoch 54/100 5/5 [==============================] - 0s 9ms/step - loss: 27.2334 - val_loss: 31.8141 Epoch 55/100 5/5 [==============================] - 0s 9ms/step - loss: 27.0429 - val_loss: 31.6203 Epoch 56/100 5/5 [==============================] - 0s 9ms/step - loss: 26.8717 - val_loss: 31.4249 Epoch 57/100 5/5 [==============================] - 0s 9ms/step - loss: 26.6938 - val_loss: 31.2326 Epoch 58/100 5/5 [==============================] - 0s 9ms/step - loss: 26.5233 - val_loss: 31.0405 Epoch 59/100 5/5 [==============================] - 0s 9ms/step - loss: 26.3455 - val_loss: 30.8534 Epoch 60/100 5/5 [==============================] - 0s 9ms/step - loss: 26.1782 - val_loss: 30.6668 Epoch 61/100 5/5 [==============================] - 0s 9ms/step - loss: 26.0079 - val_loss: 30.4833 Epoch 62/100 5/5 [==============================] - 0s 9ms/step - loss: 25.8457 - val_loss: 30.2995 Epoch 63/100 5/5 [==============================] - 0s 10ms/step - loss: 25.6685 - val_loss: 30.1246 Epoch 64/100 5/5 [==============================] - 0s 9ms/step - loss: 25.5158 - val_loss: 29.9458 Epoch 65/100 5/5 [==============================] - 0s 10ms/step - loss: 25.3546 - val_loss: 29.7691 Epoch 66/100 5/5 [==============================] - 0s 9ms/step - loss: 25.1913 - val_loss: 29.5966 Epoch 67/100 5/5 [==============================] - 0s 10ms/step - loss: 25.0354 - val_loss: 29.4251 Epoch 68/100 5/5 [==============================] - 0s 9ms/step - loss: 24.8917 - val_loss: 29.2493 Epoch 69/100 5/5 [==============================] - 0s 14ms/step - loss: 24.7363 - val_loss: 29.0766 Epoch 70/100 5/5 [==============================] - 0s 11ms/step - loss: 24.5775 - val_loss: 28.9099 Epoch 71/100 5/5 [==============================] - 0s 9ms/step - loss: 24.4226 - val_loss: 28.7477 Epoch 72/100 5/5 [==============================] - 0s 9ms/step - loss: 24.2824 - val_loss: 28.5830 Epoch 73/100 5/5 [==============================] - 0s 9ms/step - loss: 24.1359 - val_loss: 28.4203 Epoch 74/100 5/5 [==============================] - 0s 10ms/step - loss: 23.9893 - val_loss: 28.2606 Epoch 75/100 5/5 [==============================] - 0s 9ms/step - loss: 23.8457 - val_loss: 28.1032 Epoch 76/100 5/5 [==============================] - 0s 9ms/step - loss: 23.7016 - val_loss: 27.9492 Epoch 77/100 5/5 [==============================] - 0s 10ms/step - loss: 23.5698 - val_loss: 27.7930 Epoch 78/100 5/5 [==============================] - 0s 10ms/step - loss: 23.4278 - val_loss: 27.6411 Epoch 79/100 5/5 [==============================] - 0s 11ms/step - loss: 23.2910 - val_loss: 27.4914 Epoch 80/100 5/5 [==============================] - 0s 10ms/step - loss: 23.1602 - val_loss: 27.3418 Epoch 81/100 5/5 [==============================] - 0s 10ms/step - loss: 23.0240 - val_loss: 27.1959 Epoch 82/100 5/5 [==============================] - 0s 9ms/step - loss: 22.8955 - val_loss: 27.0507 Epoch 83/100 5/5 [==============================] - 0s 9ms/step - loss: 22.7679 - val_loss: 26.9064 Epoch 84/100 5/5 [==============================] - 0s 9ms/step - loss: 22.6372 - val_loss: 26.7657 Epoch 85/100 5/5 [==============================] - 0s 9ms/step - loss: 22.5143 - val_loss: 26.6253 Epoch 86/100 5/5 [==============================] - 0s 10ms/step - loss: 22.3953 - val_loss: 26.4841 Epoch 87/100 5/5 [==============================] - 0s 9ms/step - loss: 22.2664 - val_loss: 26.3482 Epoch 88/100 5/5 [==============================] - 0s 15ms/step - loss: 22.1463 - val_loss: 26.2134 Epoch 89/100 5/5 [==============================] - 0s 10ms/step - loss: 22.0291 - val_loss: 26.0794 Epoch 90/100 5/5 [==============================] - 0s 9ms/step - loss: 21.9045 - val_loss: 25.9512 Epoch 91/100 5/5 [==============================] - 0s 9ms/step - loss: 21.7937 - val_loss: 25.8216 Epoch 92/100 5/5 [==============================] - 0s 10ms/step - loss: 21.6826 - val_loss: 25.6920 Epoch 93/100 5/5 [==============================] - 0s 11ms/step - loss: 21.5674 - val_loss: 25.5655 Epoch 94/100 5/5 [==============================] - 0s 10ms/step - loss: 21.4545 - val_loss: 25.4417 Epoch 95/100 5/5 [==============================] - 0s 9ms/step - loss: 21.3462 - val_loss: 25.3186 Epoch 96/100 5/5 [==============================] - 0s 9ms/step - loss: 21.2345 - val_loss: 25.1992 Epoch 97/100 5/5 [==============================] - 0s 10ms/step - loss: 21.1315 - val_loss: 25.0790 Epoch 98/100 5/5 [==============================] - 0s 9ms/step - loss: 21.0326 - val_loss: 24.9571 Epoch 99/100 5/5 [==============================] - 0s 10ms/step - loss: 20.9255 - val_loss: 24.8388 Epoch 100/100 5/5 [==============================] - 0s 11ms/step - loss: 20.8209 - val_loss: 24.7236 2/2 [==============================] - 0s 3ms/step Mean Squared Error (MSE): 20.70429471740339 Mean Absolute Error (MAE): 3.912386627197266
import numpy as np
import matplotlib.pyplot as plt
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense, Dropout
from tensorflow.keras.callbacks import EarlyStopping
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error, mean_absolute_error
import yfinance as yf
from datetime import datetime
# List of stocks
tech_list = ['JPM', 'BAC', 'WFC', 'C']
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# Download and prepare data for each stock
data = {}
for stock in tech_list:
data[stock] = yf.download(stock, start=start, end=end)['Adj Close']
# Parameters for the LSTM model
n_steps = 60 # Size of the time series window
n_features = 1 # Number of features (only closing price is used)
# Store predictions for each stock
predictions = {}
for stock in tech_list:
print(f"\n\nTraining model for {stock}...\n")
# Prepare the data
dataset = data[stock].values.reshape(-1, 1)
scaler = MinMaxScaler(feature_range=(0, 1))
dataset = scaler.fit_transform(dataset)
# Window the time series data
X, y = [], []
dates = data[stock].index[n_steps:] # Dates for the predictions
for i in range(n_steps, len(dataset)):
X.append(dataset[i - n_steps:i, 0])
y.append(dataset[i, 0])
X, y = np.array(X), np.array(y)
# Split the data into training and test sets
train_size = int(len(X) * 0.8)
X_train, X_test = X[:train_size], X[train_size:]
y_train, y_test = y[:train_size], y[train_size:]
dates_train, dates_test = dates[:train_size], dates[train_size:]
# Reshape the input data for LSTM
X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], n_features))
X_test = X_test.reshape((X_test.shape[0], X_test.shape[1], n_features))
# Define the LSTM model
model_lstm = Sequential()
model_lstm.add(LSTM(units=50, return_sequences=True, input_shape=(n_steps, n_features)))
model_lstm.add(Dropout(0.2))
model_lstm.add(LSTM(units=50, return_sequences=False))
model_lstm.add(Dropout(0.2))
model_lstm.add(Dense(units=1))
# Compile the model
model_lstm.compile(optimizer='adam', loss='mean_squared_error')
# Early Stopping
early_stopping = EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True)
# Train the model
history = model_lstm.fit(X_train, y_train, epochs=100, batch_size=32, validation_split=0.2, callbacks=[early_stopping], verbose=1)
# Make predictions
predictions[stock] = model_lstm.predict(X_test)
# Compute performance metrics
mse = mean_squared_error(y_test, predictions[stock])
rmse = np.sqrt(mse)
mae = mean_absolute_error(y_test, predictions[stock])
print(f"{stock} - Mean Squared Error (MSE): {mse}")
print(f"{stock} - Root Mean Squared Error (RMSE): {rmse}")
print(f"{stock} - Mean Absolute Error (MAE): {mae}")
# Inverse transform the predictions and true values to get back to original scale
predictions_inverse = scaler.inverse_transform(predictions[stock])
y_test_inverse = scaler.inverse_transform(y_test.reshape(-1, 1))
# Plot the predictions vs true values with dates on x-axis
plt.figure(figsize=(14, 7))
plt.plot(dates_test, y_test_inverse, label='True Values', alpha=0.6)
plt.plot(dates_test, predictions_inverse, label='Predictions', alpha=0.6)
plt.title(f"{stock} - True vs Predicted Values")
plt.xlabel('Date')
plt.ylabel('Price')
plt.legend()
plt.grid(True)
plt.xticks(rotation=45)
plt.show()
[*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed
Training model for JPM... Epoch 1/100 4/4 [==============================] - 4s 269ms/step - loss: 0.0649 - val_loss: 0.0599 Epoch 2/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0162 - val_loss: 0.0437 Epoch 3/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0165 - val_loss: 0.0088 Epoch 4/100 4/4 [==============================] - 0s 32ms/step - loss: 0.0071 - val_loss: 0.0302 Epoch 5/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0118 - val_loss: 0.0299 Epoch 6/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0082 - val_loss: 0.0109 Epoch 7/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0059 - val_loss: 0.0112 Epoch 8/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0059 - val_loss: 0.0127 Epoch 9/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0060 - val_loss: 0.0088 Epoch 10/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0053 - val_loss: 0.0091 Epoch 11/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0042 - val_loss: 0.0094 Epoch 12/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0039 - val_loss: 0.0135 Epoch 13/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0059 - val_loss: 0.0144 2/2 [==============================] - 1s 0s/step JPM - Mean Squared Error (MSE): 0.003740853633300942 JPM - Root Mean Squared Error (RMSE): 0.06116251820601357 JPM - Mean Absolute Error (MAE): 0.047670171480103794
Training model for BAC... Epoch 1/100 4/4 [==============================] - 5s 275ms/step - loss: 0.1519 - val_loss: 0.2725 Epoch 2/100 4/4 [==============================] - 0s 44ms/step - loss: 0.0429 - val_loss: 0.0183 Epoch 3/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0193 - val_loss: 0.0106 Epoch 4/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0199 - val_loss: 0.0131 Epoch 5/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0110 - val_loss: 0.0396 Epoch 6/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0140 - val_loss: 0.0386 Epoch 7/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0108 - val_loss: 0.0178 Epoch 8/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0091 - val_loss: 0.0048 Epoch 9/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0095 - val_loss: 0.0045 Epoch 10/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0095 - val_loss: 0.0081 Epoch 11/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0066 - val_loss: 0.0140 Epoch 12/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0090 - val_loss: 0.0100 Epoch 13/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0065 - val_loss: 0.0047 Epoch 14/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0083 - val_loss: 0.0048 Epoch 15/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0068 - val_loss: 0.0063 Epoch 16/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0067 - val_loss: 0.0057 Epoch 17/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0060 - val_loss: 0.0048 Epoch 18/100 4/4 [==============================] - 0s 42ms/step - loss: 0.0066 - val_loss: 0.0048 Epoch 19/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0059 - val_loss: 0.0049 2/2 [==============================] - 1s 0s/step BAC - Mean Squared Error (MSE): 0.00207494848393367 BAC - Root Mean Squared Error (RMSE): 0.045551602429921935 BAC - Mean Absolute Error (MAE): 0.039732036632534655
Training model for WFC... Epoch 1/100 4/4 [==============================] - 4s 272ms/step - loss: 0.1176 - val_loss: 0.2333 Epoch 2/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0290 - val_loss: 0.0231 Epoch 3/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0293 - val_loss: 0.0137 Epoch 4/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0137 - val_loss: 0.0255 Epoch 5/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0160 - val_loss: 0.0479 Epoch 6/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0161 - val_loss: 0.0182 Epoch 7/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0098 - val_loss: 0.0035 Epoch 8/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0121 - val_loss: 0.0173 Epoch 9/100 4/4 [==============================] - 0s 30ms/step - loss: 0.0105 - val_loss: 0.0053 Epoch 10/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0068 - val_loss: 0.0031 Epoch 11/100 4/4 [==============================] - 0s 43ms/step - loss: 0.0078 - val_loss: 0.0032 Epoch 12/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0071 - val_loss: 0.0077 Epoch 13/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0063 - val_loss: 0.0233 Epoch 14/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0069 - val_loss: 0.0159 Epoch 15/100 4/4 [==============================] - 0s 40ms/step - loss: 0.0064 - val_loss: 0.0074 Epoch 16/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0069 - val_loss: 0.0102 Epoch 17/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0069 - val_loss: 0.0193 Epoch 18/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0077 - val_loss: 0.0213 Epoch 19/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0066 - val_loss: 0.0063 Epoch 20/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0067 - val_loss: 0.0048 2/2 [==============================] - 1s 0s/step WFC - Mean Squared Error (MSE): 0.005753101722181584 WFC - Root Mean Squared Error (RMSE): 0.07584920383353792 WFC - Mean Absolute Error (MAE): 0.06105572251113962
Training model for C... Epoch 1/100 4/4 [==============================] - 5s 281ms/step - loss: 0.1425 - val_loss: 0.2373 Epoch 2/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0323 - val_loss: 0.0064 Epoch 3/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0256 - val_loss: 0.0184 Epoch 4/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0127 - val_loss: 0.0192 Epoch 5/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0086 - val_loss: 0.0479 Epoch 6/100 4/4 [==============================] - 0s 40ms/step - loss: 0.0121 - val_loss: 0.0359 Epoch 7/100 4/4 [==============================] - 0s 43ms/step - loss: 0.0083 - val_loss: 0.0105 Epoch 8/100 4/4 [==============================] - 0s 32ms/step - loss: 0.0068 - val_loss: 0.0069 Epoch 9/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0074 - val_loss: 0.0062 Epoch 10/100 4/4 [==============================] - 0s 31ms/step - loss: 0.0051 - val_loss: 0.0099 Epoch 11/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0055 - val_loss: 0.0123 Epoch 12/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0065 - val_loss: 0.0072 Epoch 13/100 4/4 [==============================] - 0s 31ms/step - loss: 0.0053 - val_loss: 0.0069 Epoch 14/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0046 - val_loss: 0.0067 Epoch 15/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0046 - val_loss: 0.0067 Epoch 16/100 4/4 [==============================] - 0s 32ms/step - loss: 0.0059 - val_loss: 0.0069 Epoch 17/100 4/4 [==============================] - 0s 31ms/step - loss: 0.0041 - val_loss: 0.0067 Epoch 18/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0048 - val_loss: 0.0065 Epoch 19/100 4/4 [==============================] - 0s 41ms/step - loss: 0.0046 - val_loss: 0.0064 2/2 [==============================] - 1s 2ms/step C - Mean Squared Error (MSE): 0.003896662620360626 C - Root Mean Squared Error (RMSE): 0.06242325384310422 C - Mean Absolute Error (MAE): 0.05168852790376802
import numpy as np
import matplotlib.pyplot as plt
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dense, Dropout
from tensorflow.keras.callbacks import EarlyStopping
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error, mean_absolute_error
import pandas as pd
import yfinance as yf
from datetime import datetime
# Hisse senedi listesi
tech_list = ['JPM', 'BAC', 'WFC', 'C']
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# Her hisse senedi için veri indirme
data = {}
for stock in tech_list:
data[stock] = yf.download(stock, start=start, end=end)['Adj Close']
# LSTM modeli için parametreler
n_steps = 60 # Zaman serisi penceresi boyutu
n_features = 1 # Öznitelik sayısı (sadece kapanış fiyatı kullanılacak)
# Her hisse senedi için ayrı LSTM modeli ve tahminler
for stock in tech_list:
print(f"\n\nTraining model for {stock}...\n")
# Veriyi hazırlama
dataset = data[stock].values.reshape(-1, 1)
scaler = MinMaxScaler(feature_range=(0, 1))
dataset = scaler.fit_transform(dataset)
# Zaman serisi verilerini pencereleme
X, y = [], []
dates = data[stock].index[n_steps:] # Tahminlerde kullanılacak tarihler
for i in range(n_steps, len(dataset)):
X.append(dataset[i - n_steps:i, 0])
y.append(dataset[i, 0])
X, y = np.array(X), np.array(y)
# Veriyi eğitim ve test setlerine ayırma (şimdi tarihlerle birlikte)
train_size = int(len(X) * 0.8)
X_train, X_test = X[:train_size], X[train_size:]
y_train, y_test = y[:train_size], y[train_size:]
dates_train, dates_test = dates[:train_size], dates[train_size:]
# Girdi verilerini LSTM için şekillendirme
X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], n_features))
X_test = X_test.reshape((X_test.shape[0], X_test.shape[1], n_features))
# LSTM modeli tanımlama
model_lstm = Sequential()
model_lstm.add(LSTM(units=50, return_sequences=True, input_shape=(n_steps, n_features)))
model_lstm.add(Dropout(0.2))
model_lstm.add(LSTM(units=50, return_sequences=False))
model_lstm.add(Dropout(0.2))
model_lstm.add(Dense(units=1))
# Modeli derleme
model_lstm.compile(optimizer='adam', loss='mean_squared_error')
# Early Stopping
early_stopping = EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True)
# Modeli eğitme
history = model_lstm.fit(X_train, y_train, epochs=100, batch_size=32, validation_split=0.2, callbacks=[early_stopping], verbose=1)
# Tahmin yapma
predictions = model_lstm.predict(X_test)
# Tahmin performansını hesaplama
mse = mean_squared_error(y_test, predictions)
rmse = np.sqrt(mse)
mae = mean_absolute_error(y_test, predictions)
print(f"{stock} - Mean Squared Error (MSE): {mse}")
print(f"{stock} - Root Mean Squared Error (RMSE): {rmse}")
print(f"{stock} - Mean Absolute Error (MAE): {mae}")
# Tahminler ve gerçek değerleri ölçek geri dönüşümü (inverse transform)
predictions_inverse = scaler.inverse_transform(predictions)
y_test_inverse = scaler.inverse_transform(y_test.reshape(-1, 1))
# Tahminler ve gerçek değerleri görselleştirme
plt.figure(figsize=(14, 7))
plt.plot(dates_test, y_test_inverse, label='True Values', alpha=0.6)
plt.plot(dates_test, predictions_inverse, label='Predictions', alpha=0.6)
plt.title(f"{stock} - True vs Predicted Values")
plt.xlabel('Date')
plt.ylabel('Price')
plt.legend()
plt.grid(True)
plt.show()
[*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed
Training model for JPM... Epoch 1/100 4/4 [==============================] - 5s 283ms/step - loss: 0.1564 - val_loss: 0.2392 Epoch 2/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0283 - val_loss: 0.0298 Epoch 3/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0278 - val_loss: 0.0116 Epoch 4/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0104 - val_loss: 0.0420 Epoch 5/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0165 - val_loss: 0.0471 Epoch 6/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0124 - val_loss: 0.0173 Epoch 7/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0069 - val_loss: 0.0120 Epoch 8/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0095 - val_loss: 0.0132 Epoch 9/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0049 - val_loss: 0.0100 Epoch 10/100 4/4 [==============================] - 0s 40ms/step - loss: 0.0061 - val_loss: 0.0105 Epoch 11/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0050 - val_loss: 0.0115 Epoch 12/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0045 - val_loss: 0.0158 Epoch 13/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0045 - val_loss: 0.0139 Epoch 14/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0059 - val_loss: 0.0115 Epoch 15/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0046 - val_loss: 0.0108 Epoch 16/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0047 - val_loss: 0.0132 Epoch 17/100 4/4 [==============================] - 0s 32ms/step - loss: 0.0049 - val_loss: 0.0109 Epoch 18/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0048 - val_loss: 0.0100 Epoch 19/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0040 - val_loss: 0.0117 2/2 [==============================] - 1s 0s/step JPM - Mean Squared Error (MSE): 0.0035091659327544053 JPM - Root Mean Squared Error (RMSE): 0.05923821345005608 JPM - Mean Absolute Error (MAE): 0.044867550674825736
Training model for BAC... Epoch 1/100 4/4 [==============================] - 5s 473ms/step - loss: 0.1508 - val_loss: 0.2458 Epoch 2/100 4/4 [==============================] - 0s 42ms/step - loss: 0.0379 - val_loss: 0.0054 Epoch 3/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0243 - val_loss: 0.0136 Epoch 4/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0181 - val_loss: 0.0181 Epoch 5/100 4/4 [==============================] - 0s 32ms/step - loss: 0.0128 - val_loss: 0.0501 Epoch 6/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0157 - val_loss: 0.0389 Epoch 7/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0123 - val_loss: 0.0115 Epoch 8/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0088 - val_loss: 0.0045 Epoch 9/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0092 - val_loss: 0.0046 Epoch 10/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0085 - val_loss: 0.0080 Epoch 11/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0094 - val_loss: 0.0105 Epoch 12/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0075 - val_loss: 0.0061 Epoch 13/100 4/4 [==============================] - 0s 35ms/step - loss: 0.0073 - val_loss: 0.0051 Epoch 14/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0076 - val_loss: 0.0050 Epoch 15/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0065 - val_loss: 0.0056 Epoch 16/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0068 - val_loss: 0.0052 Epoch 17/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0065 - val_loss: 0.0050 Epoch 18/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0056 - val_loss: 0.0051 2/2 [==============================] - 1s 0s/step BAC - Mean Squared Error (MSE): 0.002107627561641279 BAC - Root Mean Squared Error (RMSE): 0.04590890503640093 BAC - Mean Absolute Error (MAE): 0.03998342878074972
Training model for WFC... Epoch 1/100 4/4 [==============================] - 4s 269ms/step - loss: 0.0998 - val_loss: 0.1816 Epoch 2/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0223 - val_loss: 0.0098 Epoch 3/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0235 - val_loss: 0.0095 Epoch 4/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0125 - val_loss: 0.0213 Epoch 5/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0127 - val_loss: 0.0423 Epoch 6/100 4/4 [==============================] - 0s 32ms/step - loss: 0.0145 - val_loss: 0.0186 Epoch 7/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0091 - val_loss: 0.0030 Epoch 8/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0091 - val_loss: 0.0082 Epoch 9/100 4/4 [==============================] - 0s 32ms/step - loss: 0.0088 - val_loss: 0.0034 Epoch 10/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0063 - val_loss: 0.0033 Epoch 11/100 4/4 [==============================] - 0s 46ms/step - loss: 0.0073 - val_loss: 0.0031 Epoch 12/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0072 - val_loss: 0.0087 Epoch 13/100 4/4 [==============================] - 0s 31ms/step - loss: 0.0057 - val_loss: 0.0080 Epoch 14/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0055 - val_loss: 0.0051 Epoch 15/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0067 - val_loss: 0.0061 Epoch 16/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0066 - val_loss: 0.0111 Epoch 17/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0061 - val_loss: 0.0070 2/2 [==============================] - 1s 5ms/step WFC - Mean Squared Error (MSE): 0.004144926823172298 WFC - Root Mean Squared Error (RMSE): 0.06438110610398284 WFC - Mean Absolute Error (MAE): 0.05118902110999339
Training model for C... Epoch 1/100 4/4 [==============================] - 5s 278ms/step - loss: 0.1037 - val_loss: 0.1372 Epoch 2/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0186 - val_loss: 0.0327 Epoch 3/100 4/4 [==============================] - 0s 41ms/step - loss: 0.0266 - val_loss: 0.0081 Epoch 4/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0101 - val_loss: 0.0310 Epoch 5/100 4/4 [==============================] - 0s 50ms/step - loss: 0.0114 - val_loss: 0.0470 Epoch 6/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0122 - val_loss: 0.0259 Epoch 7/100 4/4 [==============================] - 0s 39ms/step - loss: 0.0066 - val_loss: 0.0071 Epoch 8/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0075 - val_loss: 0.0058 Epoch 9/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0070 - val_loss: 0.0067 Epoch 10/100 4/4 [==============================] - 0s 34ms/step - loss: 0.0058 - val_loss: 0.0100 Epoch 11/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0052 - val_loss: 0.0090 Epoch 12/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0058 - val_loss: 0.0061 Epoch 13/100 4/4 [==============================] - 0s 41ms/step - loss: 0.0051 - val_loss: 0.0061 Epoch 14/100 4/4 [==============================] - 0s 37ms/step - loss: 0.0043 - val_loss: 0.0060 Epoch 15/100 4/4 [==============================] - 0s 36ms/step - loss: 0.0054 - val_loss: 0.0063 Epoch 16/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0053 - val_loss: 0.0061 Epoch 17/100 4/4 [==============================] - 0s 33ms/step - loss: 0.0052 - val_loss: 0.0061 Epoch 18/100 4/4 [==============================] - 0s 38ms/step - loss: 0.0046 - val_loss: 0.0061 2/2 [==============================] - 1s 8ms/step C - Mean Squared Error (MSE): 0.003379143069198468 C - Root Mean Squared Error (RMSE): 0.058130397118878074 C - Mean Absolute Error (MAE): 0.048557334822115585
# Define stock symbols and time range
tech_list = ['JPM', 'BAC', 'WFC', 'C']
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# Download stock data
company_list = [yf.download(stock, start=start, end=end)['Adj Close'] for stock in tech_list]
# Concatenate the data into a single DataFrame
closing_df = pd.concat(company_list, axis=1)
closing_df.columns = tech_list
# Prepare data for training and testing
X = closing_df.values
y = closing_df.shift(-1).fillna(method='ffill').values # Predict next day Adj Close for each stock
# Split the dataset into training and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Choose a scaler: MinMaxScaler or StandardScaler
scaler = MinMaxScaler() # or StandardScaler()
# Fit the scaler on the training data and transform both training and test data
X_train_normalized = scaler.fit_transform(X_train)
X_test_normalized = scaler.transform(X_test)
# Define the Dense Neural Network (DNN) model
input_dim = X_train_normalized.shape[1]
model_dnn = Sequential()
model_dnn.add(Dense(128, activation='relu', input_shape=(input_dim,)))
model_dnn.add(Dense(64, activation='relu'))
model_dnn.add(Dense(32, activation='relu'))
model_dnn.add(Dense(1, activation='linear')) # Output layer for regression
# Compile the DNN model
model_dnn.compile(optimizer=Adam(learning_rate=0.001), loss='mean_squared_error')
# Implement Early Stopping
early_stopping = EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True)
# Train the DNN model for each stock
predictions_dnn = []
mse_dnn = []
mae_dnn = []
rmse_dnn = []
plt.figure(figsize=(14, 10))
for i, stock in enumerate(tech_list):
# Prepare y_train for the current stock
y_train_stock = y_train[:, i]
# Train the DNN model for the current stock
history_dnn = model_dnn.fit(
X_train_normalized,
y_train_stock,
epochs=100,
batch_size=64,
validation_split=0.2,
callbacks=[early_stopping],
verbose=0
)
# Make predictions with DNN for the current stock
predictions_stock = model_dnn.predict(X_test_normalized)
# Evaluate DNN model performance for the current stock
mse_stock = mean_squared_error(y_test[:, i], predictions_stock)
mae_stock = mean_absolute_error(y_test[:, i], predictions_stock)
rmse_stock = np.sqrt(mse_stock)
# Store evaluation metrics
mse_dnn.append(mse_stock)
mae_dnn.append(mae_stock)
rmse_dnn.append(rmse_stock)
# Plot predictions vs actual values for the current stock
plt.subplot(2, 2, i+1)
plt.plot(closing_df.index[-len(predictions_stock):], y_test[:, i], label='True Values', alpha=0.6)
plt.plot(closing_df.index[-len(predictions_stock):], predictions_stock, label='DNN Predicted Values', alpha=0.6)
plt.title(f'{stock} Model: True vs Predicted Values')
plt.xlabel('Date')
plt.ylabel('Price')
plt.legend()
# Print evaluation metrics for the current stock
print(f'{stock} Model - Mean Squared Error (MSE): {mse_stock}')
print(f'{stock} Model - Mean Absolute Error (MAE): {mae_stock}')
print(f'{stock} Model - Root Mean Squared Error (RMSE): {rmse_stock}')
print()
plt.tight_layout()
plt.show()
# Print overall evaluation metrics
print('Overall Evaluation Metrics:')
for i, stock in enumerate(tech_list):
print(f'{stock} Model - Mean Squared Error (MSE): {mse_dnn[i]}')
print(f'{stock} Model - Mean Absolute Error (MAE): {mae_dnn[i]}')
print(f'{stock} Model - Root Mean Squared Error (RMSE): {rmse_dnn[i]}')
print()
[*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed
2/2 [==============================] - 0s 2ms/step JPM Model - Mean Squared Error (MSE): 282.0154147326841 JPM Model - Mean Absolute Error (MAE): 14.85076156317019 JPM Model - Root Mean Squared Error (RMSE): 16.793314584461406 2/2 [==============================] - 0s 1ms/step BAC Model - Mean Squared Error (MSE): 2.7733644737525216 BAC Model - Mean Absolute Error (MAE): 1.2851258446188534 BAC Model - Root Mean Squared Error (RMSE): 1.6653421491550984 2/2 [==============================] - 0s 991us/step WFC Model - Mean Squared Error (MSE): 1.8536676934956773 WFC Model - Mean Absolute Error (MAE): 1.0962099561504288 WFC Model - Root Mean Squared Error (RMSE): 1.361494654229563 2/2 [==============================] - 0s 2ms/step C Model - Mean Squared Error (MSE): 2.293923694494308 C Model - Mean Absolute Error (MAE): 1.2797698974609375 C Model - Root Mean Squared Error (RMSE): 1.5145704653446495
Overall Evaluation Metrics: JPM Model - Mean Squared Error (MSE): 282.0154147326841 JPM Model - Mean Absolute Error (MAE): 14.85076156317019 JPM Model - Root Mean Squared Error (RMSE): 16.793314584461406 BAC Model - Mean Squared Error (MSE): 2.7733644737525216 BAC Model - Mean Absolute Error (MAE): 1.2851258446188534 BAC Model - Root Mean Squared Error (RMSE): 1.6653421491550984 WFC Model - Mean Squared Error (MSE): 1.8536676934956773 WFC Model - Mean Absolute Error (MAE): 1.0962099561504288 WFC Model - Root Mean Squared Error (RMSE): 1.361494654229563 C Model - Mean Squared Error (MSE): 2.293923694494308 C Model - Mean Absolute Error (MAE): 1.2797698974609375 C Model - Root Mean Squared Error (RMSE): 1.5145704653446495
# Define the list of companies and their ticker symbols
company_names = ['JPMorgan', 'Bank of America', 'Wells Fargo', 'Citigroup']
tickers = ['JPM', 'BAC', 'WFC', 'C']
# Define the start and end dates for the data
end_date = datetime.now()
start_date = datetime(end_date.year - 1, end_date.month, end_date.day)
# Download stock data
data = {}
for name, ticker in zip(company_names, tickers):
data[name] = yf.download(ticker, start=start_date, end=end_date)
# Prepare the data for DNN training and prediction
predictions = {}
for name in company_names:
# Get the closing prices
closing_prices = data[name]['Adj Close'].values
# Calculate daily returns
returns = pd.Series(closing_prices).pct_change().values[1:]
# Split data into features (X) and target (y)
X = returns[:-1] # Use returns up to the second last day as features
y = returns[1:] # Predict returns for the next day
# Normalize the data
scaler = MinMaxScaler()
X_normalized = scaler.fit_transform(X.reshape(-1, 1))
y_normalized = scaler.transform(y.reshape(-1, 1))
# Split data into training and test sets
X_train, X_test, y_train, y_test = train_test_split(X_normalized, y_normalized, test_size=0.2, random_state=42)
# Define the DNN model
input_dim = X_train.shape[1]
model_dnn = Sequential()
model_dnn.add(Dense(128, activation='relu', input_shape=(input_dim,)))
model_dnn.add(Dense(64, activation='relu'))
model_dnn.add(Dense(32, activation='relu'))
model_dnn.add(Dense(1, activation='linear')) # Output layer for regression
# Compile the DNN model
model_dnn.compile(optimizer=Adam(learning_rate=0.001), loss='mean_squared_error')
# Implement Early Stopping
early_stopping = EarlyStopping(monitor='val_loss', patience=10, restore_best_weights=True)
# Train the DNN model
history_dnn = model_dnn.fit(
X_train,
y_train,
epochs=100,
batch_size=64,
validation_split=0.2,
callbacks=[early_stopping],
verbose=0
)
# Make predictions with DNN
predictions[name] = model_dnn.predict(X_test)
# Evaluate DNN model performance
mse_dnn = mean_squared_error(y_test, predictions[name])
mae_dnn = mean_absolute_error(y_test, predictions[name])
rmse_dnn = np.sqrt(mse_dnn)
print(f'{name} - Mean Squared Error (MSE): {mse_dnn}')
print(f'{name} - Mean Absolute Error (MAE): {mae_dnn}')
print(f'{name} - Root Mean Squared Error (RMSE): {rmse_dnn}')
# Plot predictions vs actual values for DNN
plt.figure(figsize=(12, 6))
plt.plot(data[name].index[-len(y_test):], y_test, label='True Values', alpha=0.6)
plt.plot(data[name].index[-len(y_test):], predictions[name], label='DNN Predicted Values', alpha=0.6)
plt.title(f'{name}: DNN True vs Predicted Values')
plt.xlabel('Date')
plt.ylabel('Value')
plt.legend()
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()
# Plot prediction distributions for DNN
plt.figure(figsize=(12, 6))
plt.hist(y_test, bins=20, alpha=0.6, label='True Values')
plt.hist(predictions[name], bins=20, alpha=0.6, label='DNN Predicted Values')
plt.title(f'{name}: Distribution of True vs Predicted Values (DNN)')
plt.xlabel('Value')
plt.ylabel('Frequency')
plt.legend()
plt.show()
[*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed [*********************100%%**********************] 1 of 1 completed
2/2 [==============================] - 0s 2ms/step JPMorgan - Mean Squared Error (MSE): 0.007951037039553136 JPMorgan - Mean Absolute Error (MAE): 0.07156918737450729 JPMorgan - Root Mean Squared Error (RMSE): 0.08916858774003958
2/2 [==============================] - 0s 2ms/step Bank of America - Mean Squared Error (MSE): 0.01676597750231491 Bank of America - Mean Absolute Error (MAE): 0.1010806614885115 Bank of America - Root Mean Squared Error (RMSE): 0.1294835028191426
2/2 [==============================] - 0s 2ms/step Wells Fargo - Mean Squared Error (MSE): 0.021392696122310612 Wells Fargo - Mean Absolute Error (MAE): 0.10618971123220296 Wells Fargo - Root Mean Squared Error (RMSE): 0.14626242211282642
2/2 [==============================] - 0s 2ms/step Citigroup - Mean Squared Error (MSE): 0.023073848574521442 Citigroup - Mean Absolute Error (MAE): 0.11331243253324136 Citigroup - Root Mean Squared Error (RMSE): 0.15190078529922563
Training Process:
High Loss Values:
Model Training and Validation Loss
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import yfinance as yf
from pandas_datareader import data as pdr
from datetime import datetime
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error, mean_absolute_error
# Download stock data
yf.pdr_override()
tech_list = ['META', 'AAPL', 'NVDA', 'NFLX']
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# Function to prepare data for each stock
def prepare_data(ticker):
stock = pdr.get_data_yahoo(ticker, start=start, end=end)
stock['10_day_MA'] = stock['Adj Close'].rolling(window=10).mean()
stock['20_day_MA'] = stock['Adj Close'].rolling(window=20).mean()
stock['50_day_MA'] = stock['Adj Close'].rolling(window=50).mean()
stock['Daily_Return'] = stock['Adj Close'].pct_change()
stock.dropna(inplace=True)
return stock
# Train and evaluate model for each stock
results = {}
for ticker in tech_list:
print(f"Processing {ticker}...")
stock = prepare_data(ticker)
# Feature selection
features = ['10_day_MA', '20_day_MA', '50_day_MA', 'Daily_Return']
X = stock[features]
y = stock['Adj Close']
# Train-test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Data scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
# Model training with hyperparameter tuning
rf = RandomForestRegressor()
param_grid = {'n_estimators': [100, 200], 'max_depth': [10, 20, None]}
grid_search = GridSearchCV(estimator=rf, param_grid=param_grid, cv=5, scoring='neg_mean_squared_error')
grid_search.fit(X_train_scaled, y_train)
# Best model
best_rf = grid_search.best_estimator_
# Predictions
y_pred = best_rf.predict(X_test_scaled)
# Evaluation
mse = mean_squared_error(y_test, y_pred)
mae = mean_absolute_error(y_test, y_pred)
# Store results
results[ticker] = {'mse': mse, 'mae': mae, 'y_test': y_test, 'y_pred': y_pred}
# Plot actual vs predicted
plt.figure(figsize=(10, 6))
plt.plot(y_test.values, label='Actual Prices')
plt.plot(y_pred, label='Predicted Prices')
plt.legend()
plt.title(f'Actual vs Predicted Stock Prices for {ticker}')
plt.xlabel('Date')
plt.ylabel('Price')
plt.show()
# Display results
for ticker, result in results.items():
print(f"{ticker}:")
print(f"Mean Squared Error: {result['mse']}")
print(f"Mean Absolute Error: {result['mae']}")
print()
[*********************100%%**********************] 1 of 1 completed
Processing META...
[*********************100%%**********************] 1 of 1 completed
Processing AAPL...
[*********************100%%**********************] 1 of 1 completed
Processing NVDA...
[*********************100%%**********************] 1 of 1 completed
Processing NFLX...
META: Mean Squared Error: 120.90425980786517 Mean Absolute Error: 6.976403434116395 AAPL: Mean Squared Error: 11.812314309833095 Mean Absolute Error: 2.6937501135105046 NVDA: Mean Squared Error: 658.4043516457613 Mean Absolute Error: 17.116642471523324 NFLX: Mean Squared Error: 196.46587527911143 Mean Absolute Error: 9.663600631341698
import pandas as pd
import numpy as np
import yfinance as yf
from datetime import datetime
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.metrics import mean_squared_error, mean_absolute_error
import matplotlib.pyplot as plt
# Define the start and end date for data download
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# List of tech stocks
tech_list = ['META', 'AAPL', 'NVDA', 'NFLX']
# Custom function to calculate RSI
def calculate_rsi(data, window=14):
delta = data.diff()
gain = (delta.where(delta > 0, 0)).rolling(window=window).mean()
loss = (-delta.where(delta < 0, 0)).rolling(window=window).mean()
rs = gain / loss
rsi = 100 - (100 / (1 + rs))
return rsi
# Custom function to calculate MACD
def calculate_macd(data, short_window=12, long_window=26, signal_window=9):
short_ema = data.ewm(span=short_window, adjust=False).mean()
long_ema = data.ewm(span=long_window, adjust=False).mean()
macd = short_ema - long_ema
signal_line = macd.ewm(span=signal_window, adjust=False).mean()
macd_hist = macd - signal_line
return macd, signal_line, macd_hist
# Function to prepare data with additional features
def prepare_data(ticker):
stock = yf.download(ticker, start=start, end=end)
stock['10_day_MA'] = stock['Adj Close'].rolling(window=10).mean()
stock['20_day_MA'] = stock['Adj Close'].rolling(window=20).mean()
stock['50_day_MA'] = stock['Adj Close'].rolling(window=50).mean()
stock['Daily_Return'] = stock['Adj Close'].pct_change()
stock['RSI'] = calculate_rsi(stock['Adj Close'])
stock['MACD'], stock['MACD_Signal'], stock['MACD_Hist'] = calculate_macd(stock['Adj Close'])
stock.dropna(inplace=True)
return stock
# Train and evaluate model for each stock
results = {}
for ticker in tech_list:
print(f"Processing {ticker}...")
stock = prepare_data(ticker)
# Feature selection
features = ['10_day_MA', '20_day_MA', '50_day_MA', 'Daily_Return', 'RSI', 'MACD', 'MACD_Signal', 'MACD_Hist']
X = stock[features]
y = stock['Adj Close']
# Train-test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Data scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
# Model training with hyperparameter tuning
gbr = GradientBoostingRegressor()
param_grid = {'n_estimators': [100, 200], 'learning_rate': [0.01, 0.1], 'max_depth': [3, 5]}
grid_search = GridSearchCV(estimator=gbr, param_grid=param_grid, cv=5, scoring='neg_mean_squared_error')
grid_search.fit(X_train_scaled, y_train)
# Best model
best_gbr = grid_search.best_estimator_
# Predictions
y_pred = best_gbr.predict(X_test_scaled)
# Evaluation
mse = mean_squared_error(y_test, y_pred)
mae = mean_absolute_error(y_test, y_pred)
# Store results
results[ticker] = {'mse': mse, 'mae': mae, 'y_test': y_test, 'y_pred': y_pred}
# Plot actual vs predicted
plt.figure(figsize=(10, 6))
plt.plot(y_test.values, label='Actual Prices')
plt.plot(y_pred, label='Predicted Prices')
plt.legend()
plt.title(f'Actual vs Predicted Stock Prices for {ticker}')
plt.xlabel('Date')
plt.ylabel('Price')
plt.show()
# Display results
for ticker, result in results.items():
print(f"{ticker}:")
print(f"Mean Squared Error: {result['mse']}")
print(f"Mean Absolute Error: {result['mae']}")
print()
[*********************100%%**********************] 1 of 1 completed
Processing META...
[*********************100%%**********************] 1 of 1 completed
Processing AAPL...
[*********************100%%**********************] 1 of 1 completed
Processing NVDA...
[*********************100%%**********************] 1 of 1 completed
Processing NFLX...
META: Mean Squared Error: 29.013272520621328 Mean Absolute Error: 3.726685070599816 AAPL: Mean Squared Error: 2.0439997239188927 Mean Absolute Error: 1.0598846605786592 NVDA: Mean Squared Error: 203.70982546183967 Mean Absolute Error: 9.570515738284836 NFLX: Mean Squared Error: 77.62173432314333 Mean Absolute Error: 6.076516550390683
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import mean_squared_error, mean_absolute_error
import yfinance as yf
import pandas as pd
import numpy as np
from datetime import datetime
import matplotlib.pyplot as plt
# Define the start and end date for data download
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# List of tech stocks
tech_list = ['META', 'AAPL', 'NVDA', 'NFLX']
# Custom function to calculate RSI
def calculate_rsi(data, window=14):
delta = data.diff()
gain = (delta.where(delta > 0, 0)).rolling(window=window).mean()
loss = (-delta.where(delta < 0, 0)).rolling(window=window).mean()
rs = gain / loss
rsi = 100 - (100 / (1 + rs))
return rsi
# Custom function to calculate MACD
def calculate_macd(data, short_window=12, long_window=26, signal_window=9):
short_ema = data.ewm(span=short_window, adjust=False).mean()
long_ema = data.ewm(span=long_window, adjust=False).mean()
macd = short_ema - long_ema
signal_line = macd.ewm(span=signal_window, adjust=False).mean()
macd_hist = macd - signal_line
return macd, signal_line, macd_hist
# Function to prepare data with additional features
def prepare_data(ticker):
stock = yf.download(ticker, start=start, end=end)
stock['10_day_MA'] = stock['Adj Close'].rolling(window=10).mean()
stock['20_day_MA'] = stock['Adj Close'].rolling(window=20).mean()
stock['50_day_MA'] = stock['Adj Close'].rolling(window=50).mean()
stock['Daily_Return'] = stock['Adj Close'].pct_change()
stock['RSI'] = calculate_rsi(stock['Adj Close'])
stock['MACD'], stock['MACD_Signal'], stock['MACD_Hist'] = calculate_macd(stock['Adj Close'])
# Adding Bollinger Bands
stock['20_day_std'] = stock['Adj Close'].rolling(window=20).std()
stock['Upper_Band'] = stock['20_day_MA'] + (stock['20_day_std'] * 2)
stock['Lower_Band'] = stock['20_day_MA'] - (stock['20_day_std'] * 2)
# Lag features
stock['Lag_1'] = stock['Adj Close'].shift(1)
stock['Lag_2'] = stock['Adj Close'].shift(2)
stock.dropna(inplace=True)
return stock
# Train and evaluate model for each stock
results = {}
for ticker in tech_list:
print(f"Processing {ticker}...")
stock = prepare_data(ticker)
# Feature selection
features = ['10_day_MA', '20_day_MA', '50_day_MA', 'Daily_Return', 'RSI', 'MACD', 'MACD_Signal', 'MACD_Hist',
'Upper_Band', 'Lower_Band', 'Lag_1', 'Lag_2']
X = stock[features]
y = stock['Adj Close']
# Train-test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Data scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
# Model training with hyperparameter tuning
gbr = GradientBoostingRegressor()
param_grid = {'n_estimators': [100, 200], 'learning_rate': [0.01, 0.1], 'max_depth': [3, 5]}
grid_search = GridSearchCV(estimator=gbr, param_grid=param_grid, cv=5, scoring='neg_mean_squared_error')
grid_search.fit(X_train_scaled, y_train)
# Best model
best_gbr = grid_search.best_estimator_
# Predictions
y_pred = best_gbr.predict(X_test_scaled)
# Evaluation
mse = mean_squared_error(y_test, y_pred)
mae = mean_absolute_error(y_test, y_pred)
# Store results
results[ticker] = {'mse': mse, 'mae': mae, 'y_test': y_test, 'y_pred': y_pred}
# Plot actual vs predicted
plt.figure(figsize=(10, 6))
plt.plot(y_test.values, label='Actual Prices')
plt.plot(y_pred, label='Predicted Prices')
plt.legend()
plt.title(f'Actual vs Predicted Stock Prices for {ticker}')
plt.xlabel('Date')
plt.ylabel('Price')
plt.show()
# Display results
for ticker, result in results.items():
print(f"{ticker}:")
print(f"Mean Squared Error: {result['mse']}")
print(f"Mean Absolute Error: {result['mae']}")
print()
[*********************100%%**********************] 1 of 1 completed
Processing META...
Processing AAPL...
[*********************100%%**********************] 1 of 1 completed
[*********************100%%**********************] 1 of 1 completed
Processing NVDA...
[*********************100%%**********************] 1 of 1 completed
Processing NFLX...
META: Mean Squared Error: 76.43674072380207 Mean Absolute Error: 4.670031624959053 AAPL: Mean Squared Error: 1.3000270080000167 Mean Absolute Error: 0.8162233908216887 NVDA: Mean Squared Error: 196.0717772692133 Mean Absolute Error: 8.60221105899563 NFLX: Mean Squared Error: 91.47280031585008 Mean Absolute Error: 5.347218156949888
import yfinance as yf
import pandas as pd
import numpy as np
from datetime import datetime
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.metrics import mean_squared_error, mean_absolute_error
import matplotlib.pyplot as plt
# Define the start and end date for data download
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# List of tech stocks
tech_list = ['META', 'AAPL', 'NVDA', 'NFLX']
# Custom function to calculate RSI
def calculate_rsi(data, window=14):
delta = data.diff()
gain = (delta.where(delta > 0, 0)).rolling(window=window).mean()
loss = (-delta.where(delta < 0, 0)).rolling(window=window).mean()
rs = gain / loss
rsi = 100 - (100 / (1 + rs))
return rsi
# Custom function to calculate MACD
def calculate_macd(data, short_window=12, long_window=26, signal_window=9):
short_ema = data.ewm(span=short_window, adjust=False).mean()
long_ema = data.ewm(span=long_window, adjust=False).mean()
macd = short_ema - long_ema
signal_line = macd.ewm(span=signal_window, adjust=False).mean()
macd_hist = macd - signal_line
return macd, signal_line, macd_hist
# Custom function to calculate On-Balance Volume (OBV)
def calculate_obv(data):
obv = (np.sign(data['Adj Close'].diff()) * data['Volume']).fillna(0).cumsum()
return obv
# Function to prepare data with additional features
def prepare_data(ticker):
stock = yf.download(ticker, start=start, end=end)
stock['10_day_MA'] = stock['Adj Close'].rolling(window=10).mean()
stock['20_day_MA'] = stock['Adj Close'].rolling(window=20).mean()
stock['50_day_MA'] = stock['Adj Close'].rolling(window=50).mean()
stock['Daily_Return'] = stock['Adj Close'].pct_change()
stock['RSI'] = calculate_rsi(stock['Adj Close'])
stock['MACD'], stock['MACD_Signal'], stock['MACD_Hist'] = calculate_macd(stock['Adj Close'])
stock['OBV'] = calculate_obv(stock)
# Adding Bollinger Bands
stock['20_day_std'] = stock['Adj Close'].rolling(window=20).std()
stock['Upper_Band'] = stock['20_day_MA'] + (stock['20_day_std'] * 2)
stock['Lower_Band'] = stock['20_day_MA'] - (stock['20_day_std'] * 2)
# Lag features
stock['Lag_1'] = stock['Adj Close'].shift(1)
stock['Lag_2'] = stock['Adj Close'].shift(2)
stock.dropna(inplace=True)
return stock
# Train and evaluate model for each stock
results = {}
for ticker in tech_list:
print(f"Processing {ticker}...")
stock = prepare_data(ticker)
# Feature selection
features = ['10_day_MA', '20_day_MA', '50_day_MA', 'Daily_Return', 'RSI', 'MACD', 'MACD_Signal', 'MACD_Hist',
'Upper_Band', 'Lower_Band', 'Lag_1', 'Lag_2', 'OBV']
X = stock[features]
y = stock['Adj Close']
# Train-test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Data scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
# Model training with hyperparameter tuning
gbr = GradientBoostingRegressor()
param_grid = {'n_estimators': [100, 200], 'learning_rate': [0.01, 0.1], 'max_depth': [3, 5]}
grid_search = GridSearchCV(estimator=gbr, param_grid=param_grid, cv=5, scoring='neg_mean_squared_error')
grid_search.fit(X_train_scaled, y_train)
# Best model
best_gbr = grid_search.best_estimator_
# Predictions
y_pred = best_gbr.predict(X_test_scaled)
# Evaluation
mse = mean_squared_error(y_test, y_pred)
mae = mean_absolute_error(y_test, y_pred)
# Store results
results[ticker] = {'mse': mse, 'mae': mae, 'y_test': y_test, 'y_pred': y_pred}
# Plot actual vs predicted
plt.figure(figsize=(10, 6))
plt.plot(y_test.values, label='Actual Prices')
plt.plot(y_pred, label='Predicted Prices')
plt.legend()
plt.title(f'Actual vs Predicted Stock Prices for {ticker}')
plt.xlabel('Date')
plt.ylabel('Price')
plt.show()
# Display results
for ticker, result in results.items():
print(f"{ticker}:")
print(f"Mean Squared Error: {result['mse']}")
print(f"Mean Absolute Error: {result['mae']}")
print()
[*********************100%%**********************] 1 of 1 completed
Processing META...
[*********************100%%**********************] 1 of 1 completed
Processing AAPL...
[*********************100%%**********************] 1 of 1 completed
Processing NVDA...
[*********************100%%**********************] 1 of 1 completed
Processing NFLX...
META: Mean Squared Error: 72.25715131919927 Mean Absolute Error: 4.449294259040038 AAPL: Mean Squared Error: 1.087594559372442 Mean Absolute Error: 0.7286874834633493 NVDA: Mean Squared Error: 274.951002466877 Mean Absolute Error: 9.056490170455486 NFLX: Mean Squared Error: 113.62666137096123 Mean Absolute Error: 5.7536296945965715
import yfinance as yf
import pandas as pd
import numpy as np
from datetime import datetime
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.metrics import mean_squared_error, mean_absolute_error
import matplotlib.pyplot as plt
# Define the start and end date for data download
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# List of tech stocks
tech_list = ['META', 'AAPL', 'NVDA', 'NFLX']
# Custom function to calculate RSI
def calculate_rsi(data, window=14):
delta = data.diff()
gain = (delta.where(delta > 0, 0)).rolling(window=window).mean()
loss = (-delta.where(delta < 0, 0)).rolling(window=window).mean()
rs = gain / loss
rsi = 100 - (100 / (1 + rs))
return rsi
# Custom function to calculate MACD
def calculate_macd(data, short_window=12, long_window=26, signal_window=9):
short_ema = data.ewm(span=short_window, adjust=False).mean()
long_ema = data.ewm(span=long_window, adjust=False).mean()
macd = short_ema - long_ema
signal_line = macd.ewm(span=signal_window, adjust=False).mean()
macd_hist = macd - signal_line
return macd, signal_line, macd_hist
# Custom function to calculate On-Balance Volume (OBV)
def calculate_obv(data):
obv = (np.sign(data['Adj Close'].diff()) * data['Volume']).fillna(0).cumsum()
return obv
# Function to prepare data with additional features
def prepare_data(ticker):
stock = yf.download(ticker, start=start, end=end)
stock['10_day_MA'] = stock['Adj Close'].rolling(window=10).mean()
stock['20_day_MA'] = stock['Adj Close'].rolling(window=20).mean()
stock['50_day_MA'] = stock['Adj Close'].rolling(window=50).mean()
stock['Daily_Return'] = stock['Adj Close'].pct_change()
stock['RSI'] = calculate_rsi(stock['Adj Close'])
stock['MACD'], stock['MACD_Signal'], stock['MACD_Hist'] = calculate_macd(stock['Adj Close'])
stock['OBV'] = calculate_obv(stock)
# Adding Bollinger Bands
stock['20_day_std'] = stock['Adj Close'].rolling(window=20).std()
stock['Upper_Band'] = stock['20_day_MA'] + (stock['20_day_std'] * 2)
stock['Lower_Band'] = stock['20_day_MA'] - (stock['20_day_std'] * 2)
# Lag features
stock['Lag_1'] = stock['Adj Close'].shift(1)
stock['Lag_2'] = stock['Adj Close'].shift(2)
stock.dropna(inplace=True)
return stock
# Train and evaluate model for each stock
results = {}
for ticker in tech_list:
print(f"Processing {ticker}...")
stock = prepare_data(ticker)
# Feature selection
features = ['10_day_MA', '20_day_MA', '50_day_MA', 'Daily_Return', 'RSI', 'MACD', 'MACD_Signal', 'MACD_Hist',
'Upper_Band', 'Lower_Band', 'Lag_1', 'Lag_2', 'OBV']
X = stock[features]
y = stock['Adj Close']
# Train-test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Data scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
# Model training with hyperparameter tuning
gbr = GradientBoostingRegressor()
param_grid = {'n_estimators': [100, 200], 'learning_rate': [0.01, 0.1], 'max_depth': [3, 5]}
grid_search = GridSearchCV(estimator=gbr, param_grid=param_grid, cv=5, scoring='neg_mean_squared_error')
grid_search.fit(X_train_scaled, y_train)
# Best model
best_gbr = grid_search.best_estimator_
# Predictions
y_pred = best_gbr.predict(X_test_scaled)
# Evaluation
mse = mean_squared_error(y_test, y_pred)
mae = mean_absolute_error(y_test, y_pred)
# Store results
results[ticker] = {'mse': mse, 'mae': mae, 'y_test': y_test, 'y_pred': y_pred}
# Plot actual vs predicted
plt.figure(figsize=(10, 6))
plt.plot(y_test.values, label='Actual Prices')
plt.plot(y_pred, label='Predicted Prices')
plt.legend()
plt.title(f'Actual vs Predicted Stock Prices for {ticker}')
plt.xlabel('Date')
plt.ylabel('Price')
plt.show()
# Display results
for ticker, result in results.items():
print(f"{ticker}:")
print(f"Mean Squared Error: {result['mse']}")
print(f"Mean Absolute Error: {result['mae']}")
print()
[*********************100%%**********************] 1 of 1 completed
Processing META...
[*********************100%%**********************] 1 of 1 completed
Processing AAPL...
[*********************100%%**********************] 1 of 1 completed
Processing NVDA...
[*********************100%%**********************] 1 of 1 completed
Processing NFLX...
META: Mean Squared Error: 73.0949546393748 Mean Absolute Error: 4.409037919228732 AAPL: Mean Squared Error: 1.0968229504555969 Mean Absolute Error: 0.7266481957246425 NVDA: Mean Squared Error: 283.5288200224573 Mean Absolute Error: 9.342451907546923 NFLX: Mean Squared Error: 108.91117071554243 Mean Absolute Error: 5.752782404856094
import pandas as pd
import numpy as np
import yfinance as yf
from datetime import datetime
from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.metrics import mean_squared_error, mean_absolute_error
import joblib
import matplotlib.pyplot as plt
# Define the start and end date for data download
end = datetime.now()
start = datetime(end.year - 1, end.month, end.day)
# Define the list of tech stocks
tech_list = ['META', 'AAPL', 'NVDA', 'NFLX']
# Function to prepare data
def prepare_data(ticker):
stock = yf.download(ticker, start=start, end=end)
stock['10_day_MA'] = stock['Adj Close'].rolling(window=10).mean()
stock['20_day_MA'] = stock['Adj Close'].rolling(window=20).mean()
stock['50_day_MA'] = stock['Adj Close'].rolling(window=50).mean()
stock['Daily_Return'] = stock['Adj Close'].pct_change()
stock.dropna(inplace=True)
X = stock[['10_day_MA', '20_day_MA', '50_day_MA', 'Daily_Return']]
y = stock['Adj Close']
return X, y
# Train and evaluate model for each stock
best_models = {}
results = {}
for ticker in tech_list:
print(f"Processing {ticker}...")
X, y = prepare_data(ticker)
# Train-test split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
# Data scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)
# Define the model
gbr = GradientBoostingRegressor()
# Define the hyperparameter grid
param_grid = {
'n_estimators': [100, 200],
'learning_rate': [0.01, 0.1],
'max_depth': [3, 5]
}
# Perform grid search cross-validation
grid_search = GridSearchCV(estimator=gbr, param_grid=param_grid, cv=5, scoring='neg_mean_squared_error')
grid_search.fit(X_train_scaled, y_train)
# Get the best model
best_model = grid_search.best_estimator_
best_models[ticker] = best_model
# Save the best model
joblib.dump(best_model, f"{ticker}_model.pkl")
# Evaluate the best model
y_pred = best_model.predict(X_test_scaled)
mse = mean_squared_error(y_test, y_pred)
mae = mean_absolute_error(y_test, y_pred)
print(f"{ticker}:")
print(f"Mean Squared Error: {mse}")
print(f"Mean Absolute Error: {mae}")
print()
# Store results
results[ticker] = {"y_test": y_test, "y_pred": y_pred}
# Select the best model based on MSE
best_ticker, best_model = min(best_models.items(), key=lambda x: mean_squared_error(x[1].predict(X_test_scaled), y_test))
print(f"The best model is for {best_ticker} with MSE: {mean_squared_error(best_model.predict(X_test_scaled), y_test)}")
# Plot actual vs predicted for all stocks
plt.figure(figsize=(12, 8))
actual_colors = ['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728']
predicted_colors = ['pink', 'purple', 'yellow', 'turquoise']
for i, (ticker, result) in enumerate(results.items()):
actual_color = actual_colors[i] if i < len(actual_colors) else actual_colors[i % len(actual_colors)] # Cycle through the color list
predicted_color = predicted_colors[i] if i < len(predicted_colors) else predicted_colors[i % len(predicted_colors)] # Cycle through the color list
plt.plot(result['y_test'].values, label=f'Actual Prices ({ticker})', color=actual_color)
plt.plot(result['y_pred'], label=f'Predicted Prices ({ticker})', linestyle='--', color=predicted_color)
plt.legend()
plt.title('Actual vs Predicted Stock Prices')
plt.xlabel('Date')
plt.ylabel('Price')
plt.show()
[*********************100%%**********************] 1 of 1 completed
Processing META...
[*********************100%%**********************] 1 of 1 completed
META: Mean Squared Error: 51.96913393722696 Mean Absolute Error: 4.891407625582708 Processing AAPL...
[*********************100%%**********************] 1 of 1 completed
AAPL: Mean Squared Error: 11.880072865328799 Mean Absolute Error: 2.651919873190455 Processing NVDA...
[*********************100%%**********************] 1 of 1 completed
NVDA: Mean Squared Error: 561.9708017720646 Mean Absolute Error: 15.392059045649653 Processing NFLX...
NFLX: Mean Squared Error: 192.7201903391627 Mean Absolute Error: 9.34184438684127 The best model is for NFLX with MSE: 192.7201903391627
